Welcome Here  Shenzhen Mingjiada Electronics Co., Ltd.

The Intel® Gaudi 2E AI accelerator now offers deep optimisation support for DeepSeek-V3.1. With outstanding performance and cost-effectiveness, Intel Gaudi 2E achieves breakthroughs in model training and real-time responsiveness in inference deployment with lower investment and higher efficiency, providing a new option for accelerating the implementation of large models.

Equipped with 96 GB of high-capacity memory and advanced HBM controllers, the Intel Gaudi 2E undergoes deep optimisation for both random and linear access scenarios. This effectively mitigates latency in AI training or inference tasks, ensuring seamless computational workflows. Boasting exceptional scalability, the Gaudi 2E supports multi-card interconnectivity, delivering flexible, customisable solutions to meet evolving AI demands.

With its outstanding adaptability and ease of use, Intel Gaudi 2E supports numerous large-scale model applications. Test data demonstrates that with Intel Gaudi 2E support, DeepSeek-V3.1 achieves significant capability enhancements in both question-answering and coding tasks: When running the DeepSeek-V3.1 model on an all-in-one server equipped with eight Intel Gaudi 2E units, under conditions of 1k input/output token length and 30 concurrent users, the concurrent token generation rate reached 10 tokens per second. Under conditions of 2k input/output token length and 28 concurrent users, the concurrent token generation rate also reached 10 tokens per second.

Whether demanding logically rigorous mathematical computations or testing analytical capabilities through knowledge comprehension, DeepSeek-V3.1 on Intel Gaudi 2E delivers swift responses and efficient parsing. This potent combination not only significantly enhances problem-solving efficiency but also empowers users to effortlessly overcome obstacles in multidimensional, high-difficulty reasoning scenarios.

Through its open architecture, robust technical support, and close ecosystem partnerships, Intel will continue to empower innovation and development within the AI industry, accelerating the widespread adoption of large AI model technologies.